1 / 15

Provisioning Other Services

Explore how to optimize service provision using Enterprise Virtualization and Cloud technologies for maximum efficiency and resilience. Discover the shift from Hyper-V to VMware and the utilization of shared storage in this comprehensive guide.

nadined
Download Presentation

Provisioning Other Services

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Provisioning Other Services Alexander Dibbo

  2. Contents • What do we mean by other services? • What services we run? • How do we run services now? • Why? • Changes that need to happen • Use of Enterprise Virtualisation and Cloud • How will we run services? • Longer term plans

  3. What do we mean by other services? • Anything which is not a critical infrastructure component of the storage or batch systems • These will continue to be run on bare metal or Enterprise Virtualisation as appropriate • Any general services we provide • Any services we host for VOs • Any services which support the rest of the infrastructure

  4. What services do we run? GeneralServices CernVM FileSystem (Stratum 0 and Stratum 1) ARGUS (Site and NGI) FTS LFC BDII (Site and Top) ARC CEs Squids Uis MyProxy VOServices Frontier VO Boxes Rucio (In testing with SKA) Infrastructure Services Database Cluster (MariaDB + Galera) Database (MariaDB or MySQL) DNS Monitoring Config Management Load Balancers

  5. How we run services now? • Multiple infrastructures on which can be hosted • Most services have individual endpoints or round robin DNS for high availability • Some services have been moved to load balancers to provide high availability and scale out capability • Some services are “Pets” with single instances

  6. How do we run services now? VMWare In full operation Can support larger VMs (CPU + RAM not storage) than HyperV Consolidation within SCD Bare Metal Necessary for some services. Can be hard to justify with squeezed budgets Hyper V Cluster In full Operation Standalone In full operation Cloud OpenNebula In full operation but winding down OpenStack Shared Storage In full operation

  7. How do we run services now? VMWare Awaiting migration of services Bare Metal CernVM File System (Stratum 1) DNS Hyper V Cluster CernVM File System (Stratum 0) DNS, Frontier DB Cluster Config Man CEs, Squids, myproxy BDII, ARGUS, FTS Standalone Frontier DB Cluster BDII, ARGUS, FTS CEs, Squids Cloud OpenNebula Frontier OpenStack Shared Storage Rucio

  8. Why? • Consolidation of expertise in SCD • There is more expertise in SCD on VMWare so that is the Enterprise Virtualisation we should run • Reduce effort required to run the services we need to run. • Cloud can be great • We should exploit the STFC Cloud where it makes sense to • Make services more resilient • Services behind load balancers can have nodes fail without bringing down the service • This should reduce call outs

  9. Changes that NEED to happen • Migration of services on Hyper V cluster to VMWare cluster • Needs to happened by end of November • Hyper V Cluster decommissioning • Shared storage will be out of warranty at the end of November • OpenNebula decommissioning • User migration is in progress. Current target is end of October. Racks are needed for next year’s procurements. • Local storage hypervisors and flavors for OpenStack • Has already been tested and will be rolled out soon. First should be in place in October. • Create a second VMWare Cluster • This will be the decommissioned hyper V cluster • All appropriate services should be behind load balancers

  10. Use of Enterprise Virtualisation • Only production services will be allowed • Limited resource pre production environments will be allowed • No development services

  11. Use of Cloud (Tier 1 perspective) • Any appropriately architected service allowed • Service owners should make sure VMs are distributed and make use of shared and local storage • Any service which is managed externally • VO services • Additional instances of any service for availability or performance reasons • Bursting of batch farm

  12. How will we run services? Bare Metal CernVM File System (Stratum 1) DNS VMWare Clusters CernVM File System (Stratum 0) DNS, Frontier DB Cluster Config Man CEs, Squids BDII, ARGUS, FTS Frontier MyProxy Cloud OpenStack Shared Storage Rucio Frontier VO Boxes Bursting of services Local Storage Rucio Frontier

  13. Longer Term • No current plans to change the operating model. • Minimise effort spent of R&D • Exploit efforts around container orchestration where appropriate • Work is being done within SCD around Kubernetes, Rancher and OpenShift (STFC Cloud Team, Data Division and DAFNI) • Other efforts may be generalisable • Exploit containerised software where easy and appropriate • If there becomes a simpler way we will use it

  14. What about Other Other Services?Facilities/Cloud/IRIS • It is expected that STFC Facilities services which work with the Tier 1 will also follow this plan • Ever increasing use of STFC Cloud for service delivery • Data-Analysis-as-a-Service • STFC Cloud team (supports SCD, STFC Facilities, ALC and IRIS) is expanding • Focus on enabling technologies (Self service virtualisation and networking, orchestration of VMs and Containers, Load balancers) • Give “users” what they need to run their own reliable services • Create Digital Assets where appropriate to support user communities • i.e. multi tenant Rucio deployment • If anyone wants to know more about these then give me a shout.

  15. Any Questions? alexander.dibbo@stfc.ac.uk

More Related